BERTAug affects Proper Noun

[memahesh]

Hi,

I have been using your BERTAug. for Text Augmentation. It works fine on a lot of tasks but it starts messing up the Proper Nouns.

Is there any fix to this ?

[makcedward]

@memahesh
There is a bug. The expected output should concatenate tokenizer word. It will be fixed in next release and the expected output should be not include "#".

For proper nouns case, it will treat as general word. In other words, "Mahesh" treats as a word and it may replace by other words. Here is one of the possible outcome.
bhesh is a good person
If you prefer to reserve the proper noun, you may leverage include it in stopwords list and passing to augmenter. Here is the example.

stopwords = ['mahesh']
aug = naw.BertAug(action=Action.SUBSTITUTE, stopwords=stopwords)

[makcedward]

Fixed in BETA release. You may try again. It will be included 0.0.4 release.

[anvitha-jain]

AttributeError: module 'nlpaug.augmenter.word' has no attribute 'BertAug'
why?